Psychology, Public Policy, and Law Copyright 2005 by the American Psychological Association 2005, Vol. 11, No. 2, 328–336 1076-8971/05/$12.00 DOI: 10.1037/1076-8971.11.2.328 WANTED: MORE RACE REALISM, LESS MORALISTIC FALLACY J. Philippe Rushton Arthur R. Jensen The University of Western Ontario University of California, Berkeley Despite repeated claims to the contrary, there has been no narrowing of the 15- to 18-point average IQ difference between Blacks and Whites (1.1 standard devia- tions); the differences are as large today as they were when first measured nearly 100 years ago. They, and the concomitant difference in standard of living, level of education, and related phenomena, lie in factors that are largely heritable, not cultural. The IQ differences are attributable to differences in brain size more than to racism, stereotype threat, item selection on tests, and all the other suggestions given by the commentators. It is time to meet reality. It is time to stop committing the “moralistic fallacy” that good science must conform to approved outcomes. In our target article (Rushton & Jensen, 2005), we proposed a hereditarian model—50% genetic–50% environmental—to explain the 15- to 18-point average IQ difference (1.1 standard deviations) between Blacks and Whites. We reviewed the worldwide distribution of test scores, the g factor of mental ability, the heritability of within- and between-groups differences, the relation of brain size to intelligence and of race differences in brain size, regression to the mean, cross- racial adoption studies, racial admixture studies, and data from life-history traits and human origins research. We were unable to identify (in Section 12 of Rushton & Jensen, 2005) any reliable environmental contribution to the Black–White IQ difference, including the non-g Flynn effect (i.e., the secular rise in IQ scores). We also found that on many dimensions, East Asian–White differences were a mirror image of Black–White differences. In Section 14, we concluded in favor of an even stronger hereditarian model—80% genetic–20% environmental—based on Jensen’s (1998, p. 443) “default hypothesis” that, by adulthood, genetic and environmental factors carry the same weight in causing group differences as they do in causing individual differences. Gottfredson (2005) is the only commentator who confronted head-on all the empirical, theoretical, and moral issues. The other commentators (Nisbett, 2005; Sternberg, 2005; Suzuki & Aronson, 2005) sidestepped the totality of the three- way race–behavior matrix shown in our Table 3. They invoked one or other of the culture-only refrains, that “race” is only “skin deep”; if not, then any difference is too small to matter; if not, then it is due to cultural factors such as statistical artifacts, insensitive tests, racism, stereotype threat, and poverty; if not, then it is poor form to talk about it. They also offered the usual culture-only promissory notes that the Black–White IQ gap can be reduced by economic improvements, interventionist programs, culture-friendly assessment systems, and nonweighted J. Philippe Rushton, Department of Psychology, The University of Western Ontario, London, Ontario, Canada; Arthur R. Jensen, School of Education, University of California, Berkeley. Correspondence concerning this article should be addressed to J. Philippe Rushton, Department of Psychology, The University of Western Ontario, London, Ontario N6A 5C2, Canada. E-mail: [email protected] 328 THE MORALISTIC FALLACY 329 models of gene–environment interaction. Their examples only confirm what we described in Sections 2, 13, and 14: Culture-only theory is a degenerating research paradigm. Are Black–White IQ Differences Narrowing? Nisbett (2005) provided the most empirically forceful of the rebuttals. He claimed that the Black–White IQ difference had decreased to only 10 points in magnitude (Ͻ0.70 standard deviations) and that it could be eliminated altogether within 20 to 60 years. He based this assertion on a purported narrowing of the Black–White difference on school achievement tests (reading, vocabulary, and mathematics), which he then extrapolated to the IQ differences. Reality, however, is stubborn. Jensen (1998, pp. 375–376, n. 33, 407–408, 494–495) showed that gains in scholastic achievement do not equal gains in g, and the Black–White differences in g are as large as ever, even for measures of reaction time. Jensen’s conclusion dovetails with a meta-analysis by Roth, Bevier, Bobko, Switzer, and Tyler (2001) that we cited at the opening of our target article. They found a mean effect size of 1.1 standard deviations that ranged from 0.38 to 1.46 (based on a sample of 6,246,729 from corporate, military, and higher education samples) depending on the g loading of the test. On the question of whether the difference was diminishing, they suggested any reduction was “either small, potentially a function of sampling error . or nonexistent for highly g loaded instruments [italics added]” (Roth et al., 2001, p. 323). In her commentary, Gottfredson (2005) underscored this message with evi- dence that no narrowing had taken place in average Black–White differences. She contrasted Black–White differences on highly g-loaded “IQ tests” with those on less g-loaded “school achievement tests.” Gottfredson found that Black–White differences on IQ tests remained constant at 1.0 standard deviation throughout the 20th century. She agreed that the differences on school achievement tests did narrow slightly from 1.07 to 0.89 standard deviations from the 1970s to the 1990s when the National Assessment of Educational Progress collected data on 9- to 17-year-olds. However, as she then pointed out, even this 20% reduction in educational achievement (a) had occurred by the mid-1980s and no longer continues, (b) is compatible with the group differences in g, and (c) does not contradict the hereditarian hypothesis. These variable Black–White differences are explained by Spearman’s (1927) hypothesis, which states that Black–White IQ differences are “most marked in just those [tests] which are known to be saturated with g” (p. 379; see Section 4 of Rushton & Jensen, 2005). The differences are lower on specific tests of memory, or arithmetic and spelling, than they are on general reasoning and transforming information. One implication is that test constructors could in principle reduce the Black–White difference to zero (or even reverse it) by including only non-g items (or those negatively loaded on g). However, they would then be left with a test that had little or no predictive power. Roth et al.’s (2001) meta-analysis con- cluded: “Overall, the results for both industrial and educational samples provide support for Spearman’s hypothesis. That is, black–white differences on measures of cognitive ability tended to increase with the saturation of g in the measure of ability” (Roth et al., 2001, p. 317). 330 RUSHTON AND JENSEN There is in fact no good evidence, contrary to Nisbett (2005; and Suzuki & Aronson, 2005), that g is malleable by nonbiological variables. That would require not just evidence that training produces higher scores but evidence of broad transfer of training effects to other highly g-loaded tasks. Extrapolation of the trends into the future may be like extrapolating the non-g secular rise in IQ scores (the Flynn effect; see Section 12). That the Flynn effect is not a Jensen effect (i.e., did not have a loading on the g factor) was corroborated by Wicherts et al. (2004). This is consistent with the lack of convergence of White and Black means across decades despite the overall rise in IQs. Two recent monographs show just how wide the achievement gap between Blacks and Whites remains. First, Thernstrom and Thernstrom (2003) compre- hensively documented the scale of the Black deficiency: For example, in reading, history, geography, and mathematics, 12th-grade Black students do not do as well as eighth-grade White students. The authors showed, moreover, that despite numerous, often well-publicized, countywide projects (such as the $2 billion program in affluent Montgomery County, Maryland, as well as the Kansas City, Missouri, school district, under judicial supervision since 1985), no plan has yet made a replicable dent in the Black–White achievement gap (despite low student– teacher ratios and computers in every classroom). Second, Ogbu (2003) studied the persistent underachievement of Black children in the well-to-do suburb of Shaker Heights, Ohio, as a result of concern raised by their (Black) parents, often highly paid professionals who had moved to the area specifically for its schools. The Black students did better than Black students elsewhere, but there were huge gaps between the Blacks and their non-Black counterparts. Instead of genetic differences in intelligence, both books offer variations on the usual culture-only explanations: poor schools, prejudice, stereotyping, low expectations, and alien- ation from White cultural domination. Nor do they consider regression to the mean (Section 9) or other genetically influenced traits that differentiate the races and affect attitudes to schoolwork (Section 10). Racial Admixture Studies: Direct Versus Indirect Evidence of Heritability Nisbett (2005) cited seven empirical studies on people of mixed race (based on self-reported ancestry, skin color, and blood groups) as “direct evidence” for the “nil” heritability of Black–White differences. He claimed these outweighed those we had presented (in Rushton & Jensen, 2005, Sections 7 and 8). It should be noted that Nisbett’s studies are peculiarly old, the mean year of publication being 1960 (median year 1966; range ϭ 1930 to 1977). Most are actually very weak and nondecisive, not having been replicated even once. Some are so old and recycled that Jensen (1973; see also 1998, pp. 478–483, 612) dealt with them 30 years ago! The blood-group studies could be repeated with better sampling and methods of analysis, but probably never will be because a more powerful tool, DNA analysis, is now available for this purpose.